1
$\begingroup$

I need help calculating P(alarm) from this setup:

picture 1

This problem is similar to that in this link:

http://www.cs.tut.fi/~elomaa/teach/AI-2012-7.pdf

However, the sum of all probabilities in that link is 1.0 as seen below:

picture 2

In the (cavity, catch, toothache) in that link, P(cavity) can be calculated by adding all values of that row and gives 0.2, by the rule of marginalization:

picture 3

My problem has a total sum of 4.0, and if I add all rows in the (alarm=T) column, I will get P(alarm) = 1.100, which is greater than 1.0. Can anyone tell me what I am doing wrong?

There is a similar question posted, but there is no answer to follow.

How to calculate Full joint probability distribution

================================================================== edit after Graham Kemp's comment

Actually, this is the problem setup. You are right. I left out some details. how can I solve for P(A) now?

picture 4

$\endgroup$
0
$\begingroup$

The setup is incorrect.   You appear to have the conditional probabilities for the events (alarm given cooking and smoke), rather than the joint probabilities.


You want. $\mathsf P(A{=}\mathsf T) ~=~ {{\mathsf P(S{=}\mathsf T,C{=}\mathsf T)~\mathsf P(A{=}\mathsf T\mid S{=}\mathsf T,C{=}\mathsf T)}+{\mathsf P(S{=}\mathsf T,C{=}\mathsf F)~\mathsf P(A{=}\mathsf T\mid S{=}\mathsf T,C{=}\mathsf F)}+{\mathsf P(S{=}\mathsf F,C{=}\mathsf T)~\mathsf P(A{=}\mathsf T\mid S{=}\mathsf F,C{=}\mathsf T)}+{\mathsf P(S{=}\mathsf F,C{=}\mathsf F)~\mathsf P(A{=}\mathsf T\mid S{=}\mathsf F,C{=}\mathsf F)}}$

You have, $\mathsf P(A{=}\mathsf T\mid S{=}\mathsf T,C{=}\mathsf T)=0.45, \mathsf P(A{=}\mathsf T\mid S{=}\mathsf T,C{=}\mathsf F)=0.15,$ et. cetera.

You are missing; $\mathsf P(S{=}\mathsf T,C{=}\mathsf T), \mathsf P(S{=}\mathsf T,C{=}\mathsf F), \mathsf P(S{=}\mathsf F, C{=}\mathsf T), \mathsf P(S{=}\mathsf F,C{=}\mathsf F)$


With the added information, you have $\mathsf P(S{=}\mathsf T)=0.27, \mathsf P(C{=}\mathsf T)=0.42$ and from the diagram, they are independent random variables, so $\mathsf P(S{=}\mathsf T, C{=}\mathsf T)=\mathsf P(S{=}\mathsf T)~\mathsf P(C{=}\mathsf T) = (0.27)(0.42) \\ \mathsf P(S{=}\mathsf T, C{=}\mathsf F)=\mathsf P(S{=}\mathsf T)~\mathsf P(C{=}\mathsf F) = (0.27)(0.58)\\\text{etc.} $

Then you now have enough information to calculate the join probability masses: $$\mathsf P(A{=}\mathsf T,S{=}\mathsf T,C{=}\mathsf T)=(0.27)(0.42)(0.45)=0.05103$$

And so forth.

$\endgroup$
  • $\begingroup$ I am using, Artificial Intelligence, a Modern Approach 3ed (Russell, Norvig), chapter 13. The superscripts (CC), (SC), etc..... are not used. Can you point me to a page / formula where that is used. I do not understand this. $\endgroup$ – CaTx Oct 20 '16 at 5:23
  • $\begingroup$ Sorry. I was using $C$ as the event of Cooking, and $C^\complement$ is the complement. I'll change it to random variables if you wish. $\endgroup$ – Graham Kemp Oct 20 '16 at 7:52
  • $\begingroup$ @CaTx Is that clearer? $\endgroup$ – Graham Kemp Oct 20 '16 at 8:00
  • $\begingroup$ I left out some details, which are now added. How can I solve for P(A) now? $\endgroup$ – CaTx Oct 20 '16 at 13:01
  • $\begingroup$ how can I computer P(A = T | S = T) from the given? $\endgroup$ – CaTx Oct 20 '16 at 13:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.